Re: PM3 slower than PM2 (fwd)

jrg@acm.org
Sat, 26 Apr 1997 12:14:46 +0100

In article <199704250115.SAA04706@server.livingston.com> megazone@livingston.com (MegaZone) writes:
> The V.42bis dictionary is 1/4 the size of say USR in the current code - on
> highly compressable data - like ping packets - this makes a big difference.
> On real world data (like the web) it doesn't.

but on bulk text transfer like email or news it does.

My only real concern, as you say you're working on the compression
dictionary size code, is why you ever didn't pick 2048 in the first
place.

ITU recommendation for V42bis, 'Guidance for Implementors', says (I've
been doing my reading, I certainly don't consider myself a modem expert):

II.1 Selection of N2, the total number of codewords.

[discussion of tradeoffs and noted behaviour]

A value for N2 of 2048 provides good compression performance across a
wide range of data types.

Anyway, hopefully this won't be an issue for much longer.

James.